Help improve this record. Suggest a correction
NVIDIA · indexed by the LabUpstream recipeStructured responses from a local Station
Explore SGLang serving, structured output and the conditions to check on a GB300 workstation.
THE SETUP
An upstream starting point
NVIDIA’s recipe covers Spark and Station with separate model guidance. It describes API serving, constrained output and prefix caching. Station entries are starting points; the Spark section has its own validation labels.
Open NVIDIA’s SGLang playbookWhat supports this record
Source reviewed 9 September 2026. UT has not reproduced this recipe. Platform guidance does not establish Dell-specific performance.
THE KNOWLEDGE AROUND IT